Fat Face zip-up cardigan
Korg DS-10
Mighty Boosh book



Hardware RAID? When I mirrored my 320GB disks it took a bit of time to fully construct the mirror - but the metadevice was available for use during the construction. I didn't have to wait, I could use it immediately.Kelloggs wrote:It took five hours to RAID-1 the buggers


Software RAID. I've had enough hardware fail that I want to be able to swap any old SATA card, or equivalent thereof, in-place if needed. (Thinking about it, I can't see any reason why one couldn't RAID over dissimilar storage technologies... but /dev/cciss and /dev/punched_card probably isn't the way to go). Besides, a worryingly-high number of HW RAID cards I've met sit happily at the RAID BIOS throughout the reconstruct, just to be able to give you a progress bar. I'd sooner have a running OS. (No doubt mis-)quoting from something I've seen elsewhere recently, "if it can't do a background rebuild, it isn't RAID, it's just backup."Dungeon-Dave wrote:Hardware RAID?Kelloggs wrote:It took five hours to RAID-1 the buggers
This, exactly. Ext3 onto the RAID, and away we go. (I just left it running overnight, anyway.) Wouldn't autodetect for ages, though, until logic dawned on me (at 2:30 this morning). The RAID modules need compiling into the kernel, but so do the SATA drivers, even if your boot device is IDE (which will probably end up SATAd and RAIDed, too)...Dungeon-Dave wrote:When I mirrored my 320GB disks it took a bit of time to fully construct the mirror - but the metadevice was available for use during the construction. I didn't have to wait, I could use it immediately.

Does he go all the way?Bane wrote:terry pratchett 2009 calendar

I was pondering /dev/random and /dev/null. Go on, you know it makes sense.Kelloggs wrote: but /dev/cciss and /dev/punched_card probably isn't the way to go).
(SATA *is* still IDE!)Kelloggs wrote:This, exactly. Ext3 onto the RAID, and away we go. (I just left it running overnight, anyway.) Wouldn't autodetect for ages, though, until logic dawned on me (at 2:30 this morning). The RAID modules need compiling into the kernel, but so do the SATA drivers, even if your boot device is IDE (which will probably end up SATAd and RAIDed, too)...

Personally, I think that's one of those more "honoured in the breech" sort of things. It's almost as if the SATA committee sat down and said, "Right, we need to keep 2.5"/3.5" disks and 5V/12V supplies, and keeping the IDE data layer would give everyone warm fuzzies. Now - what can we change?" which is why the Linux SCSI subsystem was a better fit than shoe-horning things like hot-plugging into the IDE layer. And then there's SATA hubs... or whatever they're calling them...Dungeon-Dave wrote:(SATA *is* still IDE!)
It's set up thus: fdisk /dev/sda and /dev/sdb to set two full-disk type fd partitions (AFAIK you need fd-type partitions to autodetect, and you can't fd-type a partitionless disk), then mdadm --create /dev/md0 -l 1 -n 2 /dev/sda1 /dev/sdb1, then mke2fs -j -m 0 /dev/md0, then (now it autostarts) a /dev/md0 /raid ext3 default 0 0 line in /etc/fstab.Dungeon-Dave wrote:I've not had to add SATA drivers nor the RAID modules in; seemed to work fine from that point onwards. Did you go for the partition or whole-disk option on yours?

Let's hope sogumboot wrote:Does he go all the way?Bane wrote:terry pratchett 2009 calendar

no.. but I was able to get RAID onto a partitionless pair under RH8 (as previous post) - I did it properly under FC4.Kelloggs wrote:(AFAIK you need fd-type partitions to autodetect, and you can't fd-type a partitionless disk)
You think correctly - just that I found I was able to bang in the mkreiserfs/mkfs.ext3 line almost as soon as I'd done the mdadm. I kept checking it (mdadm -D /dev/md0, or mdadm -E /dev/sda1) and although it was still building, the filesystem was perfectly useable. I expected a long wait until it was ready (like your 5 hours), but was pleasantly surprised.then mdadm --create /dev/md0 -l 1 -n 2 /dev/sda1 /dev/sdb1, then mke2fs...